Maternal Ybx1 safeguards zebrafish oocyte maturation and maternal-to-zygotic transition by repressing global translation

Before zygotic genome activation (ZGA), the transcriptionally silent early embryos depend solely on maternally provided materials for the initial stages of development, including fertilization, cleavage and preliminary body plan formation (Abrams and Mullins, 2009). This maternal control of development gradually switches to control by zygotic gene products during the maternal-to-zygotic transition (MZT), which is composed of maternal message decay and ZGA (Lee et al., 2014; Tadros and Lipshitz, 2009; Yartseva and Giraldez, 2015). Similar to pre-MZT embryos, late-stage oocytes that have completed volume expansion and material accumulation are also transcriptionally inactive (Edson et al., 2009). To drive oocyte maturation and early embryogenesis in the absence of transcription, protein synthesis needs to be fine-tuned by dedicated translational regulators targeting subsets of maternal mRNAs (Bazzini et al., 2016; Chen et al., 2013; Giraldez et al., 2006; Miao et al., 2017; Sha et al., 2017; Tadros et al., 2007; Winata et al., 2018). Interestingly, global maternal mRNA translation before the MZT also exhibits unique properties compared with translation in post-MZT embryos and non-embryonic contexts. Translational efficiencies in cleavage- and blastula-stage embryos of zebrafish and Xenopus are strongly coupled to poly(A) tail lengths. This coupling is lost in gastrulas, suggesting a developmental switch of translational control (Rabani et al., 2014; Subtelny et al., 2014). Therefore, in addition to sequence-specific RNA-binding proteins, translational regulators with less sequence specificity may also play important roles during early embryonic development.

Y-box binding proteins, including Ybx1, are a group of proteins found in bacteria, plants and animals (Darnbrough and Ford, 1981; Kohno et al., 2003; Sommerville and Ladomery, 1996; Wolffe et al., 1992). They were termed ‘Y-box binding’ because of early studies that discovered their ability to bind to Y-box DNA elements, although subsequent studies disclosed their relatively low specificity of interaction with DNA and RNA sequences (Didier et al., 1988; Kolluri et al., 1992; Wolffe, 1994; Wolffe et al., 1992). Multiple names, including YB, FRGY and MSY, have been assigned to these proteins. Vertebrate Y-box binding proteins are highly conserved in terms of sequence homology and are enriched in oocytes and spermatocytes (Bouvet and Wolffe, 1994; Darnbrough and Ford, 1981; Sommerville and Ladomery, 1996; Tafuri et al., 1993). They were identified as major components of cytoplasmic messenger ribonucleoproteins (mRNPs), with ubiquitous RNA-binding ability (Bouvet and Wolffe, 1994; Dearsly et al., 1985; Deschamps et al., 1992; Evdokimova et al., 1995; Minich et al., 1993; Murray et al., 1992). In the light of extensive studies over several decades, Y-box proteins have been linked to a wide range of nucleic acid-related processes including translational repression, RNA stabilization and transcriptional regulation (Bouvet and Wolffe, 1994; Evdokimova et al., 2001; Goodarzi et al., 2015; Holm et al., 2002; Meric et al., 1997; Ohga et al., 1998; Svitkin et al., 2009; Tafuri and Wolffe, 1993; Yu et al., 2002). Despite these results, which were mostly from in vitro and cell culture systems, the in vivo roles of Y-box proteins in developmental and physiological processes remain relatively unclear. Ybx1 knockout mice exhibit embryonic lethality and neural tube closure defects (Lu et al., 2005; Uchiumi et al., 2006). Ybx2 knockout leads to female and male infertility in mice, which has been linked to RNA instability in germ cells (Medvedev et al., 2011; Yang et al., 2005, 2007). It remains unclear to what extent the molecular mechanisms discovered in vitro contribute to Y-box protein functions in vivo.

In zebrafish, Ybx1 protein has been shown to physically associate with the dorsal localization element (DLE) of maternal sqt/ndr1 mRNA (Kumari et al., 2013), which encodes a Nodal ligand that is critical for mesendodermal induction and patterning following the midblastula transition (see reviews by Schier and Shen, 2000; De Robertis and Kuroda, 2004; Robertson, 2014). Dr Karuna Sampath’s group generated two mutant alleles of the zebrafish ybx1 locus: sa42 and sg8 (Kumari et al., 2013). Embryos derived from sa42 homozygous females, referred to as maternal mutants Mybx1^sa42, developed normally under normal incubation conditions (28.5°C), but, if incubated at 23°C, they formed a larger yolk syncytial layer by blastula stages and failed to initiate gastrulation. In contrast, Mybx1^sg8 mutants at 28.5°C showed abnormal development with aberrant cleavage from the 16-cell stage onward, much earlier than in the stage at which Mybx1_sa42 mutants at 23°C developed their phenotype, and were developmentally arrested. Through extensive studies using Mybx1^sa42 embryos, Kumari et al. propose that maternal Ybx1 is required for sqt RNA translocation, processing and translational repression (Kumari et al., 2013). Subsequently, DLEs have also been identified in the 3′UTRs of mRNAs of the Nodal antagonist genes lefty1 and lefty2, and can be bound by Ybx1, which results in translational repression of these zygotic mRNAs in zebrafish embryos (Zaucker et al., 2018). However, it remains unknown whether the function of maternal Ybx1 is restricted to the Nodal pathway components in zebrafish embryos.

In this study, we generated and analyzed zebrafish ybx1 mutants with a null allele. Although zygotic ybx1 mutants can develop into adulthood, loss of maternal Ybx1 causes severe defects in egg activation and early embryonic development. Further analyses revealed a crucial role of zebrafish Ybx1 as a general translational repressor in oocytes and early blastulas. Zebrafish Ybx1 exemplifies a regulatory module essential for translational control in early embryos, but dispensable at later stages.

To study the in vivo role of Ybx1, we employed CRISPR/Cas9 to edit the zebrafish ybx1 gene. The coding region within the first exon of ybx1 was targeted and several mutant alleles were generated. We focused on the ybx1^tsu3d5i allele that harbours a 3-base pair (bp) deletion and a 5-bp insertion (Fig. 1A). This mutation causes a premature stop codon and the truncated Ybx1^tsu3d5i protein is predicted to lose the entire cold shock domain (CSD) and C-terminal domain (Fig. 1A). Notably, the CSD is crucial for the nucleic acid-binding ability of Ybx1 (Kohno et al., 2003). Zygotic ybx1^{tsu3d5i/tsu3d5i} (Zybx1) mutants, obtained by crossing heterozygotes, showed normal morphology during embryogenesis and could become fertile adults. Maternal ybx1^{tsu3d5i/tsu3d5i} (Mybx1) mutant embryos were collected by crossing Zybx1 females with wild-type (WT) males. As revealed by quantitative reverse transcription PCR (qRT-PCR) and whole-mount in situ hybridization (WISH), maternal ybx1 transcripts in Mybx1 embryos were largely eliminated, which was presumably caused by nonsense-mediated mRNA decay (Fig. 1B,C). Given the protein truncation and mRNA loss, we believe that ybx1^tsu3d5i is a null allele.

Mybx1 mutant embryos exhibited pronounced morphological defects with complete penetrance (Fig. 1D). Opaque deposits were observed in mutant embryo yolk. In contrast to the orderly stacked blastomeres in WT embryos during the cleavage period, Mybx1 blastomeres showed abnormal shapes and placement. Mybx1 embryos failed to initiate epiboly and displayed two classes of morphological phenotypes starting from around 4 h postfertilization (hpf). Class I embryos, usually comprising 20%-35% of a batch, had a relatively clear border between the blastoderm and the yolk. In contrast, the blastoderm in Class II embryos intermingled with the yolk (Fig. 1D). Mybx1 embryos died gradually, when time-matched WT embryos underwent gastrulation and segmentation. MZybx1 embryos obtained from crosses between homozygous mutant fish were morphologically similar to Mybx1 embryos. Embryos obtained by crossing WT females with homozygous mutant males were normal. The Mybx1 defects could be substantially rescued by maternally expressed GFP-Ybx1 fusion protein in Mybx1; Tg(ef1α:GFP-ybx1) embryos, confirming the causal relationship between the defects and maternal ybx1 disruption (Fig. 1E). Taking these data together, we conclude that maternal, but not zygotic, ybx1 is indispensable for zebrafish embryonic development.

In addition to cleavage and gastrulation defects, we also found that Mybx1 embryos at the one-cell stage showed shortened chorion elevation distance that manifested the egg activation defect (Fig. 2A,B). To further analyze the activation phenotype of mutant eggs, we studied two events that are tightly associated with egg activation: cortical granule (CG) exocytosis and cytoplasmic streaming (Fuentes and Fernández, 2010; Hart, 1990; Li-Villarreal et al., 2015). We labelled CGs using fluorescein-conjugated Maclura pomifera lectin to assess CG exocytosis in activated eggs (Becker and Hart, 1999; Dosch et al., 2004). Compared with WT eggs, mutant eggs had abundantly retained CGs (Fig. 2C). We also injected rhodamine dye into WT and Mybx1 embryo yolk after fertilization to observe the cytoplasmic streaming. In WT embryos, vigorous cytoplasmic movement toward the animal pole was recorded (Movie 1). In contrast, Mybx1 embryos showed stagnant cytoplasmic streaming (Movie 2). These results conclude the egg activation defect in ybx1 mutants.

The appearance of embryonic abnormality in Mybx1 mutants immediately after fertilization suggested defective oogenesis. Thus, we sought to observe ybx1 mutant oocytes isolated from Zybx1 females. We sacrificed WT and Zybx1 female fish at 10 days post-purging (dpp), when WT oocytes mature to stages III and IV predominantly (Nair et al., 2013). Zybx1 ovaries were typically smaller than WT ones (Fig. S1A). The ybx1 mutant oocytes were capable of developing to stages III and IV, although showing relatively lower efficiency (Fig. S1B). WT and ybx1 mutant oocytes had indistinguishable appearances at stages III and IV (Fig. 2D). The translucency of ovulated WT stage-V eggs is a well-established hallmark of oocyte maturation. In contrast, prominent opaqueness was seen in ybx1 mutant eggs (Fig. 2D). In addition, WT late stage-III and stage-IV oocytes could efficiently mature to stage V under 17α,20β-dihydroxy-4-pregnen-3-one (17α,20β-DHP) treatment in vitro (Fig. 2E) (Seki et al., 2008; Xu et al., 2014). The ybx1 mutant oocytes exhibited reduced translucency after 17α,20β-DHP treatment, reflecting compromised maturation in vitro (Fig. 2E). Similarly, in vivo 17α,20β-DHP treatment of female fish is sufficient to induce oocyte maturation and ovulation (Tokumoto et al., 2011). The ovulated ybx1 mutant eggs after in vivo 17α,20β-DHP treatment also manifested the egg activation defect (Fig. S2A,B). During the oocyte maturation from stage IV to V, the major yolk proteins undergo cleavage, which leads to the translucency (Dosch et al., 2004). We determined the yolk protein cleavage level by the ratio between the higher and lower molecular weight yolk proteins (HYP and LYP), displayed by SDS-PAGE and Coomassie staining. In WT and ybx1 mutant stage-III/IV oocytes, the HYP/LYP ratios were similar. However, the ratio was more greatly decreased in WT stage-V eggs than in mutants (Fig. 2F). A similar phenomenon was also observed during in vitro maturation of oocytes (Fig. 2G). These results imply the deficiency of yolk protein cleavage during maturation of ybx1 mutant oocytes. Therefore, we conclude that Ybx1 depletion impairs oocyte maturation through stage V, which cannot be rescued by supplementing maturation hormone.

To characterize the cellular basis of morphological defects observed in Mybx1 embryos, we next analyzed some fundamental cellular structures. DAPI staining unveiled pronounced fragmentation of nuclei in Mybx1 mutants (Fig. 2H-J). The nuclear pore complex (NPC) in the nuclear envelope could be observed by immunostaining using the NPC antibody Mab414 in WT but not mutant embryos (Fig. 2J). In addition, plasma membrane structure and furrow formation were demolished in Mybx1 embryos, as indicated by membrane β-catenin immunofluorescence (Fig. 2K,L). α-Tubulin networks connecting the yolk and blastomeres were formed in WT embryos, but not in Mybx1 mutants (Fig. 2M). Taken together, these results reveal pleiotropic defects at the cellular level in Mybx1 mutants during cleavage and blastulation.

Given the severe early defects, the phenotype manifested at later stages in Mybx1 embryos may be the consequence of oogenesis and cleavage abnormalities and therefore does not confirm the postfertilization requirement of Ybx1, especially during blastulation and gastrulation. Therefore, we sought to generate Ybx1-depleted embryos with normal oogenesis and cleavage. To this end, we designed two translation-blocking morpholinos (ybx1-MO1 and ybx1-MO2) targeting ybx1 mRNA 5′-UTR to knock down ybx1. Because a lower dose of ybx1-MO2 generated morphological changes that were similar to a higher dose of ybx1-MO1, we chose to inject 4 ng of ybx1-MO2 (yMO) in subsequent analyses (Fig. S3A). Injection of yMO at the one-cell stage markedly reduced Ybx1 protein level in gastrulas (Fig. S3B). In comparison with embryos that were injected with the control MO (cMO), ybx1 morphants exhibited prominent epiboly arrest (Fig. 3A). yMO injection rendered the vast majority of embryos unable to initiate epiboly. The blastoderm of ybx1 morphants bent to form a crescent shape but failed to engulf the yolk and spread toward the vegetal pole. To confirm that the morphant phenotype was caused by ybx1 knockdown instead of non-specific effects of MO injection, we injected yMO into Tg(ef1α:GFP-ybx1) transgenic embryos, which have maternally overexpressed GFP-Ybx1 fusion protein encoded by the mRNA without the yMO targeting site. Epiboly was successfully restored in a considerable portion (34/55) of injected Tg(ef1α:GFP-ybx1) embryos, suggesting the specificity of the yMO knockdown effect (Fig. 3B).

It appeared that the number of cells in ybx1 morphants was reduced (Fig. 3A). Thus, we assessed cell proliferation by immunostaining the mitotic marker phospho-Histone H3 (pH3), which is typically positive in late G2-phase and M-phase cells (Fig. 3C) (Mendieta-Serrano et al., 2013). The total number and percentage of pH3⁺ cells were reduced in yMO-injected embryos, indicating the presence of the cell proliferation defect (Fig. 3C,D). DAPI staining revealed that morphant embryos had larger cell nuclei, which can be induced by cell cycle arrest (Fig. 3E) (Yao et al., 2017). In accordance with impaired cell proliferation, the total amount of DNA was reduced in morphants compared with time-matched control embryos during epiboly (Fig. 3F). Therefore, inhibition of postfertilization production of Ybx1 leads to cell proliferation impairment and epiboly initiation failure. These data validate the embryonic requirement of Ybx1 for normal blastulation and gastrulation.

To investigate the molecular foundation of the Mybx1 mutant phenotype, we performed RNA sequencing (RNA-seq) analysis to obtain the transcriptome profiles of WT and Mybx1 embryos at 0 hpf and 6 hpf. Ribosomal RNA depletion instead of poly(A) selection was used to enrich mRNAs in the samples, to avoid the influence of highly heterogeneous and variable poly(A) tail lengths in early embryos (Aanes et al., 2011; Subtelny et al., 2014). We first compared WT and mutant maternal transcriptomes in the one-cell-stage (0 hpf) embryos, as Y-box proteins have been shown to ubiquitously associate with maternal mRNAs in oocytes (Bouvet and Wolffe, 1994; Kohno et al., 2003). We analyzed 10,785 abundant maternal protein-coding mRNAs [>5 fragments per kilobase of transcript per million (FPKM) in either WT or Mybx1 at 0 hpf]. The majority of them showed comparable maternal expression levels in WT and mutants (84.1% with <2-fold difference) (Fig. S4A). The qRT-PCR validation ruled out the possible impact of total mRNA quantity difference (Fig. S4B,C). Therefore, global mRNA instability was not found in 0 hpf Mybx1 mutants. Considering the severe nuclear organization defect in Mybx1 embryos, we specifically examined maternal expression of DNA damage response (DDR) genes based on the 0 hpf RNA-seq data. The levels of key DDR genes are comparable in WT and Mybx1 embryos, suggesting that the maternal DDR programme is intact in mutants (Fig. S4D).

Given that the MZT progression associates with epiboly initiation (Lee et al., 2013; Zhao et al., 2017), we wondered whether the MZT is affected in Mybx1 and morphant embryos. The MZT is an evolutionarily conserved event that is constituted by maternal transcript clearance and ZGA. To assess the maternal transcript clearance, 3205 unstable maternal mRNAs with >5 FPKM at 0 hpf and >75% FPKM decrease from 0 hpf to 6 hpf in WT embryos were analyzed. Plotting of mRNA level changes revealed that maternal mRNA degradation is impaired in Mybx1 mutants. Unstable maternal mRNAs largely remained in post-MZT mutant embryos (Fig. 4A and Fig. S5A). qRT-PCR results confirmed that mRNA decay in Mybx1 mutants, as well as in yMO-injected embryos, was impaired (Fig. 4B). The stability of these mRNAs before MZT is not affected in Mybx1 embryos (Fig. S5B). Among the mRNAs examined by qRT-PCR, acadl and cd82b are targets of zygotic microRNA miR-430, a key mediator of maternal transcript clearance (Giraldez et al., 2006). We extended our analysis by a comparison with previously published high-throughput data concerning miR-430 targets (Giraldez et al., 2006). We examined 176 identified miR-430 target mRNAs (see Materials and Methods for details). Among those, 138 (78.4%) showed reduced decay (>1.5-fold difference) from 0 hpf to 6 hpf in Mybx1 mutants compared with WT embryos. Therefore, both miR-430-dependent and independent decay are affected in Ybx1-depleted embryos.

Similarly, we examined ZGA by analyzing 1818 early zygotic genes (>5 FPKM at 6 hpf and >4-fold increase from 0 hpf to 6 hpf in WT embryos). We noted that the vast majority (1516) of these genes have been shown to be upregulated (>twofold) during MZT in previously published RNA-seq datasets (Lee et al., 2013; Winata et al., 2018; Zhao et al., 2017). The expression levels of early zygotic genes were greatly downregulated in Mybx1 embryos compared with WT at 6 hpf (Fig. 4C and Fig. S5C). The defects of ZGA in both Mybx1 mutants and ybx1 morphants were further confirmed by qRT-PCR (Fig. 4D). It has been established that several waves of gene induction occur during ZGA (Lee et al., 2013). We found that 94 out of 1567 early zygotic genes with a >2-fold higher level in 6 hpf WT than in Mybx1 embryos belong to a total of 191 protein-coding first-wave genes defined by Lee et al. (2013). Therefore, different waves of gene activation are impaired by maternal Ybx1 loss. As miR-430 is zygotically expressed, the ZGA defect in Ybx1-depleted embryos could also contribute to the interruption of miR-430-dependent decay. Together, RNA-seq data and qRT-PCR validation demonstrate that the MZT process, composed of maternal mRNA decay and ZGA, is largely abolished in Ybx1-depleted embryos.

Our phenotype and transcriptome analyses have established the crucial role of maternal Ybx1. We further explored how Ybx1 exerts its function. Ybx1 has been recognized as both an RNA-binding and DNA-binding protein that is mainly involved in translational repression, RNA stabilization and transcriptional regulation (Evdokimova et al., 2001; Holm et al., 2002; Kohno et al., 2003; Svitkin et al., 2009). GFP-Ybx1 expressed in oocytes is not localized to the nucleus, suggesting the unlikely involvement of Ybx1 in maternal gene transcription (Fig. S6). Global transcript destabilization was not discovered through the transcriptome analysis in 0 hpf embryos (Fig. S4A).

Next, we sought to examine whether Ybx1 loss affects translation. Given the broad RNA-binding affinity known for Y-box proteins, we hypothesized that reporter mRNAs without specifically designed elements should be useful for analyzing translational efficiency. Thus, the GFP reporter mRNA that contains the common SV40 terminator sequence was injected into one-cell-stage WT and Mybx1 embryos together with rhodamine dye. The injected embryos were imaged under a fluorescence microscope at later stages and fluorescence levels were measured (Fig. 5A). The rhodamine dye acted as the loading control for injection. Fluorescence measurement showed that the reporter translation level was significantly higher in Mybx1 embryos (Fig. 5B,D). Similar translational upregulation was also recorded in yMO-injected embryos compared with cMO (Fig. 5C,D). Using western blotting analysis, we confirmed the increase of GFP reporter translation in Ybx1-depleted embryos (Fig. 5E,F). The qRT-PCR analysis revealed that the level of injected GFP reporter mRNA did not show a significant change from 2 hpf to 6 hpf in either WT or Mybx1 embryos (Fig. S8A). Therefore, the increased reporter translation in mutants may not be ascribed to altered mRNA stability. To demonstrate whether the translational change is restricted to the GFP reporter, we injected firefly luciferase (Fluc) reporter mRNA and recombinant Renilla luciferase (Rluc) protein to perform a similar assay (Fig. 5E). The Rluc protein served as the loading control. The luciferase assay result showed that Fluc reporter translation was also increased in Ybx1-deficient conditions (Fig. 5G). The elevated translation of two distinct mRNA reporters suggests the likelihood of an increase in overall translation level.

Next, we used SDS-PAGE and silver staining to assess global protein levels in embryos. To avoid the influence of impaired ZGA in Mybx1 mutants, we harvested embryos at 3 hpf for analysis. Western blotting of injected recombinant GFP protein served as the loading control for SDS-PAGE. We observed a remarkable increase of protein quantities in Mybx1 embryos (Fig. 5H). In order to visualize more protein bands, some bands were inevitably overdeveloped during silver staining because different proteins have vastly different quantities. We performed quantitative analysis of three regions in the gel with proper development. The result showed significantly upregulated protein levels in all three regions for Mybx1 embryos (Fig. 5I). Similarly, we also observed an elevated global protein level in ybx1 mutant eggs (Fig. S7A,B). In addition, we performed western blotting to check the endogenous protein levels of Gapdh and Khsrp, which were increased without mRNA level changes, in Mybx1 embryos (Fig. S8B-D).

Furthermore, we employed the puromycin incorporation assay to examine the translation level in embryos (Schmidt et al., 2009). The embryos were treated with puromycin, which could be incorporated into newly synthesized polypeptide chains. Western blotting detection of incorporated puromycin is the output assessed to measure the global protein synthesis rate (Schmidt et al., 2009). The puromycin-treated embryos exhibited stronger signals compared with untreated embryos or embryos treated with both puromycin and the translational blocker cycloheximide, verifying the feasibility of this approach (Fig. 5J). Stronger puromycin signals appeared for the Mybx1 samples in comparison with WT samples (Fig. 5K). Similar to silver staining, three regions were selected for quantitative analysis to avoid possible non-specific bands. The result indicated elevated protein synthesis levels in Mybx1 embryos (Fig. 5L). Taking the above data together, we conclude that Ybx1 is necessary for repressing global translation of maternal mRNAs during oocyte maturation and early embryonic development.

The endoplasmic reticulum (ER) functions as a crucial machinery for protein synthesis, modification and trafficking in eukaryotic cells. Cellular conditions that overwhelm the normal ER capacity, including massive protein misfolding and overloading, induce ER stress and the unfolded protein response (UPR), which potentially lead to cell death (Kaufman, 2002; Schröder and Kaufman, 2005). Therefore ER stress and the UPR are often associated with aberrant translational derepression (Kaufman, 2002; Miao et al., 2017). Given the global translation increase in embryos without Ybx1, we checked whether the excessive UPR also takes place. The splicing of XBP1 mRNA has been established as a common indicator of ER stress (Back et al., 2005; Yoshida et al., 2001). Visualization of RNA-seq reads mapped to the xbp1 gene showed prominent excision of the 26-bp intron in Mybx1 embryos compared with WT embryos (Fig. 6A). We confirmed, by RT-PCR, the altered splicing of xbp1 in both Mybx1 and yMO-injected embryos (Fig. 6B,C). Moreover, expression of atf3, a central mediator of the UPR (Schröder and Kaufman, 2005), was upregulated in Mybx1 and morphant embryos (Fig. 6D,E). Notably, 0 hpf Mybx1 embryos also showed increased atf3 expression, suggesting that the UPR is also triggered during oogenesis. These results indicate the existence of the excessive UPR in Ybx1-depleted embryos, which is presumably related to translational derepression and protein overloading.

The above results have established the connection between maternal Ybx1 and translational repression. It remains unknown to what extent the translational derepression caused by Ybx1 loss is responsible for the observed defects. To address this issue, we tested whether adding a distinct translational repressor into Mybx1 embryos could have a rescue effect. One ideal candidate is Eif4ebp1, the mammalian orthologue of which is known as 4E-BP1. It is an eIF4E inhibitor that negatively regulates cap-dependent translation (Pause et al., 1994). Mutations of two threonines to alanines render 4E-BP1 free of mTOR regulation and therefore constitutively active (CA) (Durand and Lykke-Andersen, 2013). We generated threonine-to-alanine substitutions at positions 34 and 43 of the zebrafish Eif4ebp1 to produce the CA form (Eif4ebp1-TTAA). We then constructed a transgenic line in which Eif4ebp1-TTAA is expressed under the control of the ef1α (eef1a1l1) promoter (Fig. S9A). Transgenic expression of Eif4ebp1-TTAA in WT background embryos led to general developmental delay without major morphological defects (Fig. S9B). This phenotype of delayed development is consistent with the role established for Eif4ebp1-TTAA in translational inhibition (Durand and Lykke-Andersen, 2013). We analyzed the Mybx1; Tg(ef1α:GFP-Eif4ebp1-TTAA) (Mybx1+TTAA) embryos, in which GFP-Eif4ebp1-TTAA was expressed in the absence of Ybx1 protein (Fig. 7A). Significantly rescued egg activation of Mybx1+TTAA embryos could be observed after fertilization, indicated by increased chorion elevation (Fig. 7B-D) and promoted cytoplasmic streaming (Movie 3) in contrast to Mybx1 mutants (Movie 2). Eif4ebp1-TTAA also improved the cleavage of major yolk proteins in eggs and in response to in vitro maturation (Fig. 7E,F). Importantly, a considerable portion of Mybx1+TTAA embryos were able to initiate epiboly, which was almost never seen in Mybx1 mutants (Fig. 7G,H). The MZT process was also partially recovered in epiboly restored Mybx1+TTAA embryos, evidenced by decreased unstable maternal mRNAs and increased zygotically expressed mRNAs in post-MZT embryos (Fig. 7I,J). These results together suggest that Eif4ebp1-TTAA, as a translational repressor, confers a prominent rescue effect on Mybx1 mutants. Therefore, translational repression constitutes a major part, if not all, of Ybx1 function in zebrafish embryos.

Y-box proteins are integral components of maternal mRNP particles found in Xenopus oocytes (Darnbrough and Ford, 1981; Dearsly et al., 1985; Deschamps et al., 1992; Murray et al., 1992). Because of the lack of a suitable antibody for immunostaining endogenous zebrafish Ybx1, we tried to look at subcellular localization of overexpressed HA-Ybx1 fusion protein in embryos. Interestingly, HA-Ybx1 proteins formed cytoplasmic foci when expressed in zebrafish embryos at 4 hpf (Fig. 8A,E). These structures were reminiscent of processing bodies (P-bodies), which are cytoplasmic protein aggregates involved in translational control and mRNA degradation (Decker and Parker, 2012). Previous studies have also reported the association of YBX1 with P-body components and its localization to RNA granules (Lyons et al., 2016; Somasekharan et al., 2015; Yang and Bloch, 2007). Therefore, we examined whether P-body components could also be found in Ybx1-containing foci in embryos. mRNAs encoding HA-Ybx1 and GFP-labelled P-body components Dcp1a or Patl1 were co-injected into one-cell-stage embryos. The imaging result showed that HA-Ybx1 and GFP-Dcp1a/Patl1 indeed colocalized in aggregates in embryonic cells at 4 hpf (Fig. 8A-H). Furthermore, endogenous Ybx1 can physically associate with GFP-Dcp1a, as revealed by co-immunoprecipitation (co-IP) in 8 hpf embryo lysates (Fig. 8I). Probably because of the limited protein product levels of injected mRNAs in earlier embryos, we were unable to detect colocalization and co-IP at pre-MZT stages. Overall, our results suggest that Ybx1 proteins may associate with P-body components and reside in P-body-like granules in early embryos.

We next asked whether the translational repression mediated by Ybx1 depends on its binding to individual mRNAs. The MS2-MCP system is employed to generate protein-RNA interaction (Bertrand et al., 1998). The multiple MS2 stem loops attached to specific RNA molecules are able to associate with designed MCP-tagged protein partners. We injected GFP-24×MS2 mRNA, mCherry mRNA and blue-coloured Alexa Fluor 350 (AF350) dye into WT and Mybx1 embryos. In line with the above results, both GFP and mCherry reporter translation levels were upregulated in Mybx1 embryos compared with WT embryos (Fig. 8J,K). Additional input of HA-tagged tandem dimer MCP (HA-tdMCP), which possesses GFP-24×MS2 mRNA binding ability, did not change either GFP or mCherry translation level in Mybx1 embryos. In contrast, injection of the mRNA encoding HA-tdMCP-Ybx1, which directs Ybx1 to GFP-24×MS2 mRNA, led to a significant decrease of the GFP reporter level. The mCherry level was decreased to a lesser extent, which might result from MCP-independent binding of HA-tdMCP-Ybx1 (Fig. 8J,K). These results suggest that tethering Ybx1 in proximity to the target mRNAs leads to promoted translational repression.

Ybx1, along with other Y-box binding proteins, has a long and complicated history of studies, shown by multiple names coined by different researchers. Because of its broad nucleic acid binding ability, it has been linked to various interacting partners and cellular processes. However, the in vivo roles of Y-box proteins, particularly in developmental processes, have not been well characterized using genetically modified animal models. In this study, we analyzed zebrafish Ybx1 function during early development using Mybx1 null mutants. Our phenotypic characterization unveiled pleiotropic defects caused by the loss of maternal Ybx1, including impaired oocyte maturation, deficient egg activation, cellular malformations and abolition of the MZT. Considering the severe cellular defects, crucial cell activities, including chromosome segregation and cytoskeleton-mediated processes, may be largely affected in Mybx1 mutants. Furthermore, arrest of cell cycle progression and epiboly movement in ybx1 morphants implied the continuous requirement of Ybx1 protein supply during blastulation. In Ybx1-depleted embryos, we found aberrant translational upregulation and the excessive UPR, indicating the translational repression mediated by Ybx1. The amelioration of Mybx1 defects by Eif4ebp1-TTAA validated the important role of Ybx1 as a translational repressor. We further showed the association of Ybx1 with P-body components, which are involved in mRNA silencing and decay. Taken together, these results delineate the mechanisms underlying the crucial role of maternal Ybx1 (Fig. 9).

One area of interest is whether epiboly failure and MZT loss in Mybx1 mutants are caused by defective egg activation and cleavage. Given that knockdown of ybx1 in WT embryos also impaired epiboly and the MZT without an influence on early cleavage (Fig. 3A), it is likely that later defects in Mybx1 mutants are not necessarily secondary to the early defects. However, we cannot exclude the possibility that early defects aggravate later defects in Mybx1 mutants.

The Mybx1^tsu3d5i maternal mutant embryos we generated in this report exhibit anomalous cleavage and impaired epiboly, and deform during gastrulation or early segmentation (Fig. 1D).This is similar to the phenotypic appearance of Mybx1^sg8 mutant embryos reported by Kumari et al. (2013). However, careful examination indicates that Mybx1^tsu3d5i oocytes undergo defective maturation with impaired yolk protein cleavage from stages IV to V (Fig. 2F,G) and the resulting fertilized eggs show abnormal egg activation (Fig. 2C) and cytoplasmic streaming (Movie 2). Our results support an idea that maternal Ybx1 protein starts to function before fertilization and is essential for ensuring egg quality. Any phenotypic differences among existing Mybx1^tsu3d5i, Mybx1^sa42 and Mybx1^sg8 mutants, if there are, might be explained by the different molecular nature of the mutations and/or the genetic backgrounds.

We uncovered the localization of overexpressed HA-Ybx1 fusion protein in P-body-like granules in zebrafish embryos (Fig. 8A-H). These observed granules may be homologous structures of stored mRNP particles in Xenopus oocytes. Both kinds of structures contain Y-box proteins and P-body components, and are functionally involved in maternal mRNA translational repression (Dearsly et al., 1985; Ladomery et al., 1997; Sommerville and Ladomery, 1996). Similar granules, for example the neuronal P-body-like granules, were also discovered in other types of cells (Barbee et al., 2006). Given the dispensability of zygotic ybx1 in zebrafish somatic tissues, our findings raise the possibility that Y-box proteins are functionally required components of maternal mRNA granules, but not somatic granules. In agreement with this hypothesis, colocalization and physical association of Ybx1 with overexpressed P-body components were detected in post-MZT embryos. However, the results were not achieved with pre-MZT embryos, most likely because of relatively low levels of overexpressed proteins before MZT. Future efforts are needed to verify the existence of endogenous Ybx1-containing granules in oocytes and early embryos using optimized antibodies. Interestingly, as is seen with some other RNA granule proteins, YBX1 is able to form polymerized amyloid-like fibrils in vitro (Guryanov et al., 2012; Kato et al., 2012). It remains to be seen whether Ybx1 can also heavily polymerize in oocytes and embryos.

We noticed that Eif4ebp1-TTAA fails to fully rescue the Mybx1 mutant phenotype (Fig. 7G). Therefore, we cannot exclude other potential pathways of Ybx1 functions in vertebrate oocytes and embryos (Fig. 9). For example, Ybx1 has been shown to regulate the processing of non-coding RNAs (Shurtleff et al., 2016, 2017; Wu et al., 2015). In the present study, we focused on protein-coding mRNAs in our RNA-seq analysis. It will be interesting to profile non-coding RNAs in Mybx1 mutants. Another possibility underlying the partial rescue effect is that inappropriate levels of Eif4ebp1-TTAA may impact the rescue efficiency. Nevertheless, the pronounced rescue effect of Eif4ebp1-TTAA in Mybx1 embryos establishes translational derepression as a primary cause of the severe phenotype.

The translational control is of particular importance in transcriptionally inactive early embryos, because of the lack of gene regulation at the transcription layer. Previous studies have provided numerous examples of how sequence-specific regulators shape translational levels of mRNAs containing targeted cis-regulatory elements (Chen et al., 2013; Sha et al., 2017; Tadros et al., 2007). In this study, we show that a relatively ubiquitous control mechanism also exists in oocytes and embryos to modulate translation. Furthermore, translation is globally activated following fertilization to support early embryogenesis (Bazzini et al., 2016). We demonstrate that maternal Ybx1 acts as a brake for this process and prevents protein overloading in embryos. Cooperation among diverse translational regulators ensures a proper protein synthesis level that carries the embryo through the MZT (Bazzini et al., 2016, 2012; Winata et al., 2018). Interestingly the MZT process itself, which is considered an event centred on RNA turnover, is also related to the switch of translational control modes (Schier, 2007; Subtelny et al., 2014; Winata et al., 2018). Our study highlights the presence of the Ybx1-dependent translational control module that is crucial for early embryonic development, but not required for later stages. Prospective mechanisms that also distinguish between early embryos and somatic cells will complete our knowledge of dynamic regulatory strategies during development.

Zebrafish husbandry and manipulation were approved by Tsinghua University Animal Care and Use Committee. Tuebingen strain zebrafish were used as WT fish and for generating genetically modified lines. Embryos were maintained in Holtfreter's solution at 28.5°C and staged according to the standard protocol (Kimmel et al., 1995).

To observe oocyte phenotype, ovaries were isolated from sacrificed female fish. Oocytes were immersed in oocyte culture medium [90% Leibovitz's L-15 medium (Gibco), 0.5 mg/ml bovine serum albumin (Amresco), pH 9] for imaging. Oocyte maturation in vitro was achieved following a previously established protocol (Xu et al., 2014). Oocytes were incubated in oocyte maturation medium containing 17α,20β-DHP (Sigma-Aldrich) for 3 h before imaging. For hormone-induced oocyte maturation and ovulation in vivo, female fish were placed in 1 μg/ml 17α,20β-DHP for 4 h. Ovulated eggs were squeezed out by gently pressing the female fish abdomen.

CRISPR/Cas9-mediated mutagenesis was performed as previously reported (Chang et al., 2013; Yao et al., 2017). The guide RNA (gRNA) target site was selected with CHOPCHOP (Montague et al., 2014). The PCR primers used for T7 endonuclease I assay (ybx1-target-F, ybx1-target-R) and genotyping (ybx1-wt-F, ybx1-mut-F, ybx1-target-R) are listed in Table S1. Tol2-mediated transgenesis was performed using the previously described system (Han et al., 2011). A 347 bp promoter between −326 and +21 positions from the zebrafish eef1a1l1 transcription start site was employed to drive transgene expression. For the generation of Tg(ef1α:GFP-ybx1) and Tg(ef1α:GFP-Eif4ebp1-TTAA) strains, the construct and Tol2 mRNA were injected into WT embryos. For the generation of Mybx1; Tg(ef1α:GFP-ybx1) and Mybx1; Tg(ef1α:GFP-Eif4ebp1-TTAA) strains, the construct and mRNA were injected into embryos from crosses between homozygous mutant males and heterozygous females. The identified founder fish were crossed to obtain F1 fish for producing transgenic embryos.

Open reading frames were inserted into the pCS2+ vector for in vitro synthesis of mRNAs using the mMESSAGE mMACHINE SP6 kit (Thermo Fisher Scientific). The coding sequences for 24×MS2 and tdMCP are from Addgene plasmids #31865 and #40649 (deposited by Robert Singer), respectively (Bertrand et al., 1998; Wu et al., 2012). mRNA injection doses per embryo in this study are as follows: GFP, 100 pg; Fluc, 200 pg; HA-ybx1, 200 pg; GFP-dcp1a, 400 pg; GFP-patl1a, 500 pg; GFP-24×MS2, 300 pg; mCherry, 200 pg; HA-tdMCP, 200 pg; and HA-tdMCP-ybx1, 450 pg. Rhodamine dye (Sigma-Aldrich, R8881) was injected as the loading control or to visualize cytoplasmic movement.

The antisense morpholinos were purchased from Gene Tools. Sequences are as follows: ybx1-MO1, CTCTCTAGTGTGTTTTCCCGGTGCT; ybx1-MO2 (yMO), CGGCCTCGCTGCTCATGTTGTTTTC; control MO (cMO), TAATTTACTTACCCTCAAGTTGCTG.

Recombinant protein expression and purification were performed as previously described (Liu et al., 2013). Sequences encoding Rluc, GFP and mCherry were cloned into the pGEX-6P-1 vector for GST-labelled protein expression. Transformed BL21-competent Escherichia coli cells were cultured and Glutathione Sepharose 4B (GE Healthcare) was used for protein purification. The loading control was a 50 pg injection of Rluc, GFP or mCherry protein.

Whole-mount in situ hybridization and immunofluorescence were performed following standard protocols as previously described (Liu et al., 2013; Xu et al., 2014). The digoxigenin-labelled antisense probe was synthesized using the DIG RNA labelling kit (Roche). Anti-Digoxigenin-AP antibody (Roche, 11093274910, 1:3000) and BM Purple AP substrate (Roche) were used for hybridization signal detection. Primary antibodies that were used for immunofluorescence were: anti-NPC proteins (Abcam, ab24609, 1:1000), anti-β-catenin (Sigma-Aldrich, C2206, 1:1000), anti-α-tubulin (Sigma-Aldrich, T6199, 1:400), anti-pH3 (Cell Signaling Technology, 9701, 1:200), anti-HA (Santa Cruz Biotechnology, sc-7392, 1:200) and anti-GFP (Abcam, ab13970, 1:500). Secondary antibodies were: Alexa Fluor 488 goat anti-mouse IgG (Jackson ImmunoResearch, 115-545-003, 1:200), Alexa Fluor 488 goat anti-rabbit IgG (Jackson ImmunoResearch, 111-545-003, 1:200), Alexa Fluor 647 goat anti-mouse IgG (Jackson ImmunoResearch, 115-605-003, 1:200), and Alexa Fluor 488 goat anti-chicken IgY (Abcam, ab150169, 1:200). Cortical granules were labelled using fluorescein-conjugated Maclura pomifera lectin (Vector Laboratories, FL-1341) as previously reported (Becker and Hart, 1999).

For morphological phenotype capture, embryos were maintained in Holtfreter's solution and photographed using a Nikon SMZ1500 stereomicroscope. For images showing WISH results, stained embryos were immersed in glycerol and photographed under the stereomicroscope. Live embryos and oocytes were photographed under an Olympus MVX10 fluorescence microscope for epifluorescent images. Confocal images were acquired using a Zeiss 710 or Zeiss 710 META confocal microscope. ImageJ software (http://imagej.nih.gov/ij/) was used for image analyses, including fluorescence intensity measurement, length measurement, area measurement and cell counting. When assessing chorion elevation distances, embryo and chorion diameters were first measured. The chorion elevation was calculated by 0.5×(chorion diameter−embryo diameter).

Total RNA was extracted from embryo samples using TRIzol reagent (Thermo Fisher Scientific). RNA was treated with DNase I (Tiangen Biotech) and further purified with RNeasy Mini kit (Qiagen). cDNA was generated by reverse-transcribing 1 μg of total RNA with M-MLV reverse transcriptase (Promega) and random hexamers (Promega). Quantitative PCR was performed using a Roche LightCycler 480 machine and analyzed following the 2^−ΔΔCt method. Unless otherwise noted, vasa was used as the reference gene for qRT-PCR because of its approximately equal expression levels in WT and mutants and relative stability in early embryos. Primers that were used for qRT-PCR and RT-PCR experiments are listed in Table S2. Genomic DNA quantification of embryos was performed in reference to the previously reported method (Zhao et al., 2017). Luciferase reporter assay for embryo lysates was carried out as previously described for mammalian cell samples (Liu et al., 2013).

Total RNA extraction and genomic DNA clearance were carried out as stated above. To further purify mRNA, samples were treated with NEBNext rRNA depletion kit (New England Biolabs) to eliminate rRNA. Sequencing libraries were constructed with NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs). Clustered library preparations were sequenced on an Illumina HiSeq 2500 machine and 50 bp single-end reads were generated. Clean reads, with low quality reads removed from the raw data, were mapped to the zebrafish GRCz10 reference genome using TopHat v2.0.12 (Trapnell et al., 2009). HTSeq v0.6.1 was employed to count the reads mapped to individual genes (Anders et al., 2015). FPKM values were then calculated to determine gene expression levels. Gene type information was acquired from Ensembl BioMart (http://www.ensembl.org/biomart) and only protein-coding genes were included for the transcriptome analysis in this study.

The previously published datasets were obtained from NCBI Gene Expression Omnibus for analyses. The miR-430 targets we analyzed are those mRNAs that have stable Ensembl gene IDs, contain miR-430 target sequence GCACUU in their 3′-UTRs, and are upregulated >1.5-fold in post-MZT MZdicer embryos compared with WT and miR-430-injected MZdicer embryos (Giraldez et al., 2006). For ZGA analysis, out of 269 first-wave genes that have been reported by Lee et al. (2013), we focused on 191 protein-coding ones according to BioMart gene information (Lee et al., 2013).

SDS-PAGE, silver staining and Coomassie staining were performed following the established protocol (Schägger, 2006). For each sample of the yolk protein cleavage assay, ten eggs were dechorionated and lysed in 500 μl TNE buffer [10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1% Triton X-100]. An aliquot of 5-10 μl lysate was loaded for SDS-PAGE and Coomassie staining. For puromycin incorporation assay, dechorionated embryos were treated with 50 μg/ml puromycin (Solarbio) from 2 hpf to 3 hpf. When necessary, cycloheximide was added 30 min before puromycin administration and remained at a concentration of 50 μg/ml until embryo lysis. Embryos were lysed at 3 hpf in TNE buffer and subjected to western blotting. Intensity measurement was carried out using ImageJ.

Western blotting and immunoprecipitation were carried out essentially as previously described (Liu et al., 2013). For detection of puromycin incorporation, ECL plus super sensitive chemiluminescent substrate (Solarbio) was used. The following antibodies acted as primary antibodies for western blotting: anti-YBX1 (Sigma-Aldrich, WH0004904M1, 1:500), anti-GAPDH (ZSGB-Biotech, TA-08, 1:2000), anti-KHSRP (Sigma-Aldrich, SAB2101240, 1:500), anti-puromycin (Merck Millipore, MABE434, 1:5000), anti-GFP (Santa Cruz Biotechnology, sc-9996, 1:1000), anti-mCherry (Easy Bio, BE2026, 1:1000). For co-IP, injected embryos were lysed at 8 hpf in TNE buffer for further procedures. Anti-GFP (Santa Cruz Biotechnology, sc-9996, 1:100) antibody was used for immunoprecipitation.

GraphPad Prism 7 was used for making statistical graphs and analyses. Wilcoxon rank sum tests were performed manually in Microsoft Excel. Error bars represent mean±s.d. unless otherwise noted. Sample sizes and statistical methods are specified in figure legends. All Student's and Welch's t-tests are unpaired, and all P values are two-sided.

We thank other members of the Meng laboratory for discussions, Robert Singer and Qiang Wang for sharing reagents, and the staff at the Center of Biomedical Analysis of Tsinghua University for technical assistance.